Linear Convergence of ISTA and FISTA

In this paper, we revisit the class of iterative shrinkage-thresholding algorithms (ISTA) for solving the linear inverse problem with sparse representation, which arises in signal and image processing. It is shown in the numerical experiment to deblur an image that the convergence behavior in the logarithmic-scale ordinate tends to be linear instead of logarithmic, approximating to be flat. Making meticulous observations, we find that the previous assumption for the smooth part to be convex weakens the least-square model. Specifically, assuming the smooth part to be strongly convex is more reasonable for the least-square model, even though the image matrix is probably ill-conditioned. Furthermore, we improve the pivotal inequality tighter for composite optimization with the smooth part to be strongly convex instead of general convex, which is first found in [Li et al., 2022]. Based on this pivotal inequality, we generalize the linear convergence to composite optimization in both the objective value and the squared proximal subgradient norm. Meanwhile, we set a simple ill-conditioned matrix which is easy to compute the singular values instead of the original blur matrix. The new numerical experiment shows the proximal generalization of Nesterov's accelerated gradient descent (NAG) for the strongly convex function has a faster linear convergence rate than ISTA. Based on the tighter pivotal inequality, we also generalize the faster linear convergence rate to composite optimization, in both the objective value and the squared proximal subgradient norm, by taking advantage of the well-constructed Lyapunov function with a slight modification and the phase-space representation based on the high-resolution differential equation framework from the implicit-velocity scheme.

Revisiting the acceleration phenomenon via high-resolution differential equations

Nesterov's accelerated gradient descent (NAG) is one of the milestones in the history of first-order algorithms. It was not successfully uncovered until the high-resolution differential equation framework was proposed in [Shi et al., 2022] that the mechanism behind the acceleration phenomenon is due to the gradient correction term. To deepen our understanding of the high-resolution differential equation framework on the convergence rate, we continue to investigate NAG for the $\mu$-strongly convex function based on the techniques of Lyapunov analysis and phase-space representation in this paper. First, we revisit the proof from the gradient-correction scheme. Similar to [Chen et al., 2022], the straightforward calculation simplifies the proof extremely and enlarges the step size to $s=1/L$ with minor modification. Meanwhile, the way of constructing Lyapunov functions is principled. Furthermore, we also investigate NAG from the implicit-velocity scheme. Due to the difference in the velocity iterates, we find that the Lyapunov function is constructed from the implicit-velocity scheme without the additional term and the calculation of iterative difference becomes simpler. Together with the optimal step size obtained, the high-resolution differential equation framework from the implicit-velocity scheme of NAG is perfect and outperforms the gradient-correction scheme.

LibSignal: An Open Library for Traffic Signal Control

This paper introduces a library for cross-simulator comparison of reinforcement learning models in traffic signal control tasks. This library is developed to implement recent state-of-the-art reinforcement learning models with extensible interfaces and unified cross-simulator evaluation metrics. It supports commonly-used simulators in traffic signal control tasks, including Simulation of Urban MObility(SUMO) and CityFlow, and multiple benchmark datasets for fair comparisons. We conducted experiments to validate our implementation of the models and to calibrate the simulators so that the experiments from one simulator could be referential to the other. Based on the validated models and calibrated environments, this paper compares and reports the performance of current state-of-the-art RL algorithms across different datasets and simulators. This is the first time that these methods have been compared fairly under the same datasets with different simulators.

Proximal Subgradient Norm Minimization of ISTA and FISTA

For first-order smooth optimization, the research on the acceleration phenomenon has a long-time history. Until recently, the mechanism leading to acceleration was not successfully uncovered by the gradient correction term and its equivalent implicit-velocity form. Furthermore, based on the high-resolution differential equation framework with the corresponding emerging techniques, phase-space representation and Lyapunov function, the squared gradient norm of Nesterov's accelerated gradient descent (\texttt{NAG}) method at an inverse cubic rate is discovered. However, this result cannot be directly generalized to composite optimization widely used in practice, e.g., the linear inverse problem with sparse representation. In this paper, we meticulously observe a pivotal inequality used in composite optimization about the step size $s$ and the Lipschitz constant $L$ and find that it can be improved tighter. We apply the tighter inequality discovered in the well-constructed Lyapunov function and then obtain the proximal subgradient norm minimization by the phase-space representation, regardless of gradient-correction or implicit-velocity. Furthermore, we demonstrate that the squared proximal subgradient norm for the class of iterative shrinkage-thresholding algorithms (ISTA) converges at an inverse square rate, and the squared proximal subgradient norm for the class of faster iterative shrinkage-thresholding algorithms (FISTA) is accelerated to convergence at an inverse cubic rate.

Gradient Norm Minimization of Nesterov Acceleration: $o(1/k^3)$

In the history of first-order algorithms, Nesterov's accelerated gradient descent (NAG) is one of the milestones. However, the cause of the acceleration has been a mystery for a long time. It has not been revealed with the existence of gradient correction until the high-resolution differential equation framework proposed in [Shi et al., 2021]. In this paper, we continue to investigate the acceleration phenomenon. First, we provide a significantly simplified proof based on precise observation and a tighter inequality for $L$-smooth functions. Then, a new implicit-velocity high-resolution differential equation framework, as well as the corresponding implicit-velocity version of phase-space representation and Lyapunov function, is proposed to investigate the convergence behavior of the iterative sequence $\{x_k\}_{k=0}^{\infty}$ of NAG. Furthermore, from two kinds of phase-space representations, we find that the role played by gradient correction is equivalent to that by velocity included implicitly in the gradient, where the only difference comes from the iterative sequence $\{y_{k}\}_{k=0}^{\infty}$ replaced by $\{x_k\}_{k=0}^{\infty}$. Finally, for the open question of whether the gradient norm minimization of NAG has a faster rate $o(1/k^3)$, we figure out a positive answer with its proof. Meanwhile, a faster rate of objective value minimization $o(1/k^2)$ is shown for the case $r > 2$.

An Adjoint-Free Algorithm for CNOP via Sampling

In this paper, we propose a sampling algorithm based on statistical machine learning to obtain conditional nonlinear optimal perturbation (CNOP), which is essentially different from the traditional deterministic optimization methods. The new approach does not only reduce the extremely expensive gradient (first-order) information directly by the objective value (zeroth-order) information, but also avoid the use of adjoint technique that gives rise to the huge storage problem and the instability from linearization. Meanwhile, an intuitive anlysis and a rigorous concentration inequality for the approximate gradient by sampling are shown. The numerical experiments to obtain the CNOPs by the performance of standard spatial sturctures for a theoretical model, Burgers equation with small viscosity, demonstrate that at the cost of losing accuracy, fewer samples spend time relatively shorter than the adjoint-based method and directly from definition. Finally, we reveal that the nonlinear time evolution of the CNOPs obtained by all the algorithms are almost consistent with the quantity of norm square of perturbations, their difference and relative difference on the basis of the definition method.

Modeling Network-level Traffic Flow Transitions on Sparse Data

Modeling how network-level traffic flow changes in the urban environment is useful for decision-making in transportation, public safety and urban planning. The traffic flow system can be viewed as a dynamic process that transits between states (e.g., traffic volumes on each road segment) over time. In the real-world traffic system with traffic operation actions like traffic signal control or reversible lane changing, the system's state is influenced by both the historical states and the actions of traffic operations. In this paper, we consider the problem of modeling network-level traffic flow under a real-world setting, where the available data is sparse (i.e., only part of the traffic system is observed). We present DTIGNN, an approach that can predict network-level traffic flows from sparse data. DTIGNN models the traffic system as a dynamic graph influenced by traffic signals, learns the transition models grounded by fundamental transition equations from transportation, and predicts future traffic states with imputation in the process. Through comprehensive experiments, we demonstrate that our method outperforms state-of-the-art methods and can better support decision-making in transportation.

On the Hyperparameters in Stochastic Gradient Descent with Momentum

Following the same routine as [SSJ20], we continue to present the theoretical analysis for stochastic gradient descent with momentum (SGD with momentum) in this paper. Differently, for SGD with momentum, we demonstrate it is the two hyperparameters together, the learning rate and the momentum coefficient, that play the significant role for the linear rate of convergence in non-convex optimization. Our analysis is based on the use of a hyperparameters-dependent stochastic differential equation (hp-dependent SDE) that serves as a continuous surrogate for SGD with momentum. Similarly, we establish the linear convergence for the continuous-time formulation of SGD with momentum and obtain an explicit expression for the optimal linear rate by analyzing the spectrum of the Kramers-Fokker-Planck operator. By comparison, we demonstrate how the optimal linear rate of convergence and the final gap for SGD only about the learning rate varies with the momentum coefficient increasing from zero to one when the momentum is introduced. Then, we propose a mathematical interpretation why the SGD with momentum converges faster and more robust about the learning rate than the standard SGD in practice. Finally, we show the Nesterov momentum under the existence of noise has no essential difference with the standard momentum.

On Learning Rates and Schrödinger Operators

The learning rate is perhaps the single most important parameter in the training of neural networks and, more broadly, in stochastic (nonconvex) optimization. Accordingly, there are numerous effective, but poorly understood, techniques for tuning the learning rate, including learning rate decay, which starts with a large initial learning rate that is gradually decreased. In this paper, we present a general theoretical analysis of the effect of the learning rate in stochastic gradient descent (SGD). Our analysis is based on the use of a learning-rate-dependent stochastic differential equation (lr-dependent SDE) that serves as a surrogate for SGD. For a broad class of objective functions, we establish a linear rate of convergence for this continuous-time formulation of SGD, highlighting the fundamental importance of the learning rate in SGD, and contrasting to gradient descent and stochastic gradient Langevin dynamics. Moreover, we obtain an explicit expression for the optimal linear rate by analyzing the spectrum of the Witten-Laplacian, a special case of the Schr\"odinger operator associated with the lr-dependent SDE. Strikingly, this expression clearly reveals the dependence of the linear convergence rate on the learning rate -- the linear rate decreases rapidly to zero as the learning rate tends to zero for a broad class of nonconvex functions, whereas it stays constant for strongly convex functions. Based on this sharp distinction between nonconvex and convex problems, we provide a mathematical interpretation of the benefits of using learning rate decay for nonconvex optimization.